Here, I want to simulate scRNA-seq data from trajctories to benchmark performance of velocity based visualization.
Create multifurcating trajectory
## [1] "finding approximate nearest neighbors ..."
## [1] "calculating clustering ..."
## [1] "graph modularity: 0.683635778301471"
## [1] "identifying cluster membership ..."
## com
## 1 2 3 4 5
## 154 197 246 196 207
Assign pseudotime to cells
Order cells wrt to pseudotime - subset 1 –> observed, subset 2 (at later pseudtime points) –> projected
## [1] "Done finding neighbors"
## [1] "Done making graph"
Try fdg on observed
Try removing subsets and compare velocity fdg vs observed fdg
## [1] "Done finding neighbors"
## [1] "Done making graph"
Now look at graph made from just observed
#####Scratch
Use the reticulate package to use scVelo from within R:
Extract spliced and unspliced data
Extract PCA coordinates
Filter genes
Downsample cells to make things easier
Normalize for dimensional reduction
Dimensional reduction
Run velocyto on panc data
Scores of observed and projected states in PC space
Graph visualization on subset of cells from PC coordinates
First, we’ll see if removing Ngn3 low EP cell types affects the visualization. Given that there are only relatively few of these cells, I suspect that the effect won’t be noticeable in the visualization.
As expected, the visualization doesn’t change very much by removing Ngn3 low EP cells. Next, let’s see the effect of removing Ngn3 high EP or Pre-endocrine cells.
First, remove pre-endocrine cells..
…and now Ngn3 high EP
Let’s try removing multiple subsets, Ngn3 highEP and Pre-endocrine
Remove a proportion of cells from one of the terminal cell types..